Method and apparatus for normalization and deconvolution of assay data

Abstract

The present invention is directed to deconvolution and normalization of assay data. The present invention includes a control and analysis system, used in conjunction with a signal generation and detection apparatus, for capturing, processing and analyzing images of samples having resonance light scattering (RLS) particle labels. The control and analysis system processes instructions and algorithms for performing multiplexed assays of two or more colors, for example, to allow separation and analysis of detected light that contains information from two or more different types or sizes of RLS particles. The multiplexing analysis software is preferably incorporated within the system of the present invention, and the multiplexing analysis is preferably performed in real-time during a scanning or assay procedure. The invention provides for a computer readable medium containing instructions for carrying out the same.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS[0001]This application claims priority to provisional U.S. applications, Ser. No. 60 / 317,543, filed on Sep. 5, 2001, entitled “Apparatus for Analyte Assays”, Ser. No. 60 / 364,962, filed Mar. 12, 2002, entitled “Multiplexed Assays Using Resonance Light Scattering Particles,” and Ser. No. 60 / 376,049, filed Apr. 24, 2002, entitled “Signal Generation and Detection System for Analyte Assays,” all of which are incorporated herein by reference in their entirety.FIELD OF THE INVENTION[0002]The present invention generally relates to methods of processing data obtained from assay measurements on analytes. Specifically the present invention provides methods and apparatus for correlating measured light intensity with analyte concentration and for normalization of data across microarray samples. The invention is of particular applicability to assays that use resonance light scattering particles.BACKGROUND OF THE INVENTION[0003]Binding-pair techniques play an...

AI Technical Summary

Benefits of technology

[0040]In particular, use of metal particles, particularly particles made of silver or gold, permits high levels of sensitivity, thereby permitting detection of very low quantities of particles per unit area. With the use of certain combinations of particles and methods of illumination and detection, it is possible to detect a wide range of particle densities from about 0.001 to about 103 particles per square micron (μ2) in a sample as a function of a particle size. By using appropriate type(s) of particles, different types of analytes in the same sample can be detected to very low levels and across very wide concentration ranges, as occur for example, in microarrays.

[0041]Use of RLS particles also permits the quantitative detection of single or multiple analytes in a sample and has found particularly forceful application to microarray based assay systems. The high sensitivity and ease of use of the signal generation and detection system of RLS analyte detection means that one skilled in the art can by inexpensive means, detect and measure one or more analytes in a sample to extremely low concentrations without need of signal or target analyte molecule amplification methods.

[0042]Furthermore, if a sample is to be analyzed for two or more different analytes, different types of analytes exist at different concentrations in the samples. In particular, when using solid-phase related means such as array chips, some analytes may be at higher or lower concentrations from two to a few orders of magnitude more or less than other analytes in the sample. The methods of the present invention can be used to detect different analytes that are present at different concentrations in a sample by using different particle labels, each with a different inherent sensitivity based upon its light scattering power. For example, one could use a particle of greater light scattering power to detect an analyte present at a lower concentration and vice versa. This approach permits performing assays for analytes across a broad concentration range within the same signal generation and detection system.

[0043]The appropriate measurements can be accomplished on a single apparatus by utilizing both particle counting (at low particle densities) and integrated light intensity measurements (at high densities) on the same sample. Even wider detection ranges and greater sensitivities are possible if more powerful light sources such as lasers are used, and if more sophisticated detection methods such as confocal imaging are added to basic illumination and detection methods.

Problems solved by technology

However, those methods suffer from a number of drawbacks, which makes the detection of analytes complicated, difficult, time consuming, and costly.

Not least of these drawbacks are problems of interference of chemical or enzymatic reactions, contamination, complicated and multi-step work-up and analysis procedures, limited adaptability to single step “homogeneous”, non-separation, formats, and the requirement of costly and sophisticated instrumentation.

Nevertheless, detection of RLS particles with conventional laser scanners entails some technical limitations because the particles emit the same wavelength light as the excitation source.

The principal limitation is that any incident light entering the optics on the detection side is indistinguishable from the scattered light signal because the wavelengths of the scattered light and the laser light are the same.

However, Earth observations largely consist of unknown materials whose spectra are ambiguous.

Accurate determination of material abundance from these observations is difficult.

This situation compounds the overall problem because the solution becomes under-determined since there are fewer reference materials than there are materials in the image.

However, attempts to “unmix” pixels into these component parts were only partially successful, probably due to insufficient knowledge of the full range of spectral characteristics of the target objects.

Spectral signature variability is another factor which may seriously limit the success unmixing procedures.

The high dimensionality of the hyperspectral data is a serious challenge to the operational use of the technology.

A basic problem in expression analysis is the determination of gene regulation profiles while compensating for unassociated assay and system variations.

However, independent assay and system variations conspire to obstruct accuracy in such comparative expression studies.

4. Amplification: clones are subject to PCR amplification, which is difficult to quantify and may fail completely.

In particular, there are inherent difficulties in accurately correlating a measured signal with the quantity of analyte that gives rise to the signal.

There is also no guarantee of reproducibility, i.e., that equivalently prepared microarray features (such as spots) produce equivalently quantified expression values.

Furthermore, there are problems of consistency between samples, both different samples in the same array and between samples on different arrays.

Thus, quantitative analyte detection, which is so vital to areas such as gene expression analysis, is often confounded by an inability to accurately gauge analyte presence in any given sample and to make reliable comparisons between different samples.

The drawback of existing normalization methods that do not use controls is that an analytic scheme must reverse-engineer a data model that itself attempts to accurately describe the underlying data.

However, without controls at one's disposal to help reverse the effects of systematic and protocol induced variance, it may never be possible to have a sound model.

Nevertheless, existing normalization methods that use controls are still imperfect.

Acids Res., 28: e47, (2000), described the use of averaging as an approach to normalization of data across an array, but did not provide a closed solution.

However, by using control data averages, there is still no guarantee that the normalized set of data will be characterized reproducibly in a consistent manner because all the control features are assumed to express equivalently.

Unless this is reliably true all the time, normalization performance cannot be relied up on to guarantee a consistent solution.

This method therefore suffers from the same drawback as that of Schuchhardt et al.

This method suffers from the drawback that intensity data may not obey a linear relationship with respect to analyte concentration and also that removing data points is unsatisfactory because of the associated loss of information.

Accordingly, the art lacks a normalization method that consistently produces the same effect and guarantees that equivalently prepared microarray features will produce equivalently quantified expression values across the entire range of control features.

Images